Add SiliconFlow transcription provider

[shuimaoiko]

Summary

• Add a SiliconFlow transcription provider for the official /v1/audio/transcriptions endpoint.
• Use SILICONFLOW_API_KEY for auth and FunAudioLLM/SenseVoiceSmall as the default model.
• Add aliases for sensevoice, sensevoice-small, teleai, and telespeech.
• Keep the SiliconFlow request body aligned to the documented file + model form fields.
• Fix process_file so user-owned audio that is already in the provider preferred format is not deleted as temporary converted audio.

Validation

• .venv/bin/python -m pytest tests/providers/test_siliconflow.py tests/test_registry.py tests/test_process.py -q
• .venv/bin/python -m pytest -q
• .venv/bin/python -m black --check sapat/providers/siliconflow.py sapat/process.py tests/providers/test_siliconflow.py tests/test_process.py
• .venv/bin/python -m compileall sapat tests/providers/test_siliconflow.py tests/test_process.py
• git diff --check

Companion contribution for daytonaio/content#13. No secrets, private media, transcripts, or payout details are included.